Predictors of Compliance in Mass Drug Administration for the Treatment and Prevention of Lymphatic Filariasis in Leogane, Haiti

Transcription

1 Am. J. Trop. Med. Hyg., 78(2), 2008, pp Copyright 2008 by The American Society of Tropical Medicine and Hygiene Predictors of Compliance in Mass Drug Administration for the Treatment and Prevention of Lymphatic Filariasis in Leogane, Haiti Jeffrey T. Talbot, Abigail Viall, Abdel Direny, Madsen Beau de Rochars, David Addiss, Thomas Streit, Els Mathieu, and Patrick J. Lammie* Department of Epidemiology, Rollins School of Public Health, Emory University, Atlanta, Georgia; Division of Parasitic Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia; Hopital Ste. Croix, Leogane, Haiti; Center for Tropical Disease Research and Training, University of Notre Dame, Notre Dame, Indiana Abstract. The global strategy for the elimination of lymphatic filariasis (LF) is based on annual mass drug administration (MDA) to interrupt transmission. Noncompliance with MDA represents a serious programmatic obstacle for the LF program because systematically noncompliant individuals may serve as a reservoir for the parasite and permit recrudescence of infection. Using a survey questionnaire concerning practices, beliefs, and attitudes towards MDA, we assessed differences between noncompliant individuals and compliant individuals in Leogane, Haiti (n 367) after four years of treatment. A logistic regression model showed the odds of being noncompliant were significantly increased for women (odds ratio 2.74, 95% confidence interval ), as well as for people who lacked knowledge about both LF and programs to eliminate infection. Public health programs should be designed to target people who are at risk for systematic noncompliance. INTRODUCTION Lymphatic filariasis (LF) is a mosquito-borne parasitic infection that is endemic in more than 80 countries and spans tropical areas including parts of the Americas. 1 Currently, it is estimated that more than one billion people worldwide are at risk of infection and approximately 120 million harbor the parasite. 2,3 Clinical manifestations of the disease include lymphedema and, in males, hydrocele. Lymphatic filariasis has been recognized as a leading cause of global long-term disability and a factor that contributes to poverty in filariasisendemic countries. 4 The current strategy for elimination of LF focuses on annual mass drug administration (MDA) with albendazole and diethylcarbamazine (DEC) or ivermectin. It is estimated that to interrupt transmission, drug coverage in five consecutive MDAs must exceed 65 75%. 5 7 In collaboration with the Ministry of Health, the Centers for Disease Control and Prevention (CDC), and the University of Notre Dame, a pilot program for the treatment of LF in the commune of Leogane, Haiti was developed and administered by the local hospital (Ste. Croix). 8 The first MDA occurred in October 2000 and subsequent MDAs took place annually thereafter. Albendazole and DEC were offered at drug distribution posts that were accessible to the community and included homes, churches, schools, and health clinics. During the year leading up to MDA, efforts were made to educate and mobilize the public through health educators who disseminated information about LF at community meetings. Although coverage levels varied during the first three years of operation, a substantial reduction in the prevalence of infection was observed. 8 Initial results after three rounds of MDA were promising, showing a significant decrease in the prevalence and intensity of microfilaremia by blood smear. Surveys completed after the third MDA indicated that 18% of persons more than 14 years of age had not been treated in any of the first three rounds of MDA. 9 * Address correspondence to Patrick J. Lammie, Division of Parasitic Diseases, Centers for Disease Control and Prevention, Mailstop F-36, 4770 Buford Highway, Atlanta, GA pjl1@cdc.gov 283 When a proportion of the population fails to participate in MDA, a potential reservoir for the parasite is left untreated, opening the door to recrudescence of microfilaremia and reducing the probability of successful elimination of transmission. 10,11 The purpose of this study was to investigate potential risk factors for noncompliance with MDA in Leogane, Haiti. If persons who fail to participate in MDA demonstrate common characteristics, then it may be possible to target high-risk groups with new health education messages. To date, only a few studies have focused on identifying variables that play a role in noncompliance in mass treatment programs. Thus, the results of the present study may be relevant to geographic localities and programs beyond Haiti. MATERIALS AND METHODS Selection of study population and study design. Four communities were selected in the Leogane Commune as spot check sites for intensive monitoring and evaluation of the pilot program to eliminate filariasis implemented by the Ste. Croix Hospital. Bire, Bonyotte, Guinebeau and the Rue la Liberte neighborhood were selected because they represented the geographic and ecologic diversity of filariasis in Leogane and had not been the site of previous surveys or sentinel site work related to the program. Field work in Leogane was reviewed and approved by the CDC Institutional Review Board and the Ethics Committee of Hopital Ste. Croix. An initial household census was conducted to obtain demographic information concerning the population that included age, sex, and family size. Family members who were present were allowed to answer questions concerning compliance for people in the household who were absent at the time that the initial questionnaire was administered. Noncompliance with the MDA program was found to be approximately 30% of persons surveyed. However, some of these persons had moved into Leogane after the MDA of In addition, questions regarding compliance with MDA were often answered by proxies who often were unable to accurately characterize the compliance of other family members. Microfilaremia and antigenemia were assessed in the spot check sites as described previously for the sentinel sites. 8 Persons tested

2 284 TALBOT AND OTHERS represented a convenience sample of persons residing in each community. On the basis of the compliance data gathered during the census, a study was designed to validate the extent of noncompliance and the underlying attitudes of systematically noncompliant individuals (never treated) in relation to people who had participated in MDA all four years (systematically compliant). A detailed survey was administered to determine if there were factors that could predict noncompliance, as well as to verify the number of times individuals had taken the pills. The questionnaire included a large number of inquiries that can be broadly broken into the following categories: feelings about one s role in the community, trust in the healthcare system, knowledge of the LF program and filarial disease, access to MDA, perceived personal benefit of MDA participation, perceived risk of contracting disease, and fear of side effects from taking drugs. Additionally, questions specifically asked about how many times individuals had participated in MDA. Power calculations necessary for unmatched case control studies and temporal constraints inherent in the project led to a decision to sample 440 persons from the initial census. Excluded from the cohort formation were people less than 14 years of age or those who had lived in Leogane for less than four years. Systematic random sampling was then implemented to select 220 compliant (treated all 4 years) persons and 220 noncompliant (never treated) persons for survey. Two line listings were generated, stratifying by compliance status as ascertained from the original census, and people were randomly selected based on generated sample intervals. To reduce the possibility of sampling from within the same household, the listings were ordered first by community and then by household within the community. Interviews were conducted by trained personnel who were not affiliated with the LF program or Ste. Croix Hospital, and all information was recorded on personal digital assistants to ensure a standard quality of data entry. Subjects living in the same residence were administered the questionnaire simultaneously and in isolation of each other by different interviewers to control for the sharing of information and to preserve independent sampling. After the surveys were conducted, subjects were reclassified into three groups on the basis of self-reported participation in MDA and the number of times they had taken pills. Systematically noncompliant individuals were designated cases; persons who took the pills all four times were used as controls. Persons who took the pills one to three times were considered semicompliant. Definition of main exposures and outcome variables. Based on a review of the available literature as well as unpublished data from the LF elimination program in Haiti, age, sex, and level of education were selected as descriptive main exposure variables. Other exposure variables included fear of side effects, questions concerning knowledge of LF, and a person s perceived role in the community. Age was divided into three categories; 14 24, 25 49, and 50 years of age. Persons years of age were selected as the referent group. Similarly, education was categorized by highest level completed: secondary school or higher (the referent group), primary schooling, or no education completed. For sex, males were the referent group. This was done because the Ministry of Health excluded women of reproductive age from albendazole treatment in the program s first two years of operation. As a result, men were more likely to participate in the initial rounds of MDA. 9 Effect modification by community of residence was considered. However, residence was not found to be a significant interaction term. Statistical analysis. Unless otherwise specified, all data analysis was performed using SAS version 9.1 (SAS Institute Inc., Cary, NC) and a P value 0.05 was used to determine statistical significance of associations. Chi-square analysis was conducted to evaluate the relationship between all potential exposure variables and the outcome of interest (compliance). These crude results served as the primary mechanism for determining which exposure variables seemed important and which were not, and were used to screen out insignificant exposure variables. Some exposure variables had less than 10 persons per stratum. To obtain meaningful odds ratios (ORs), Don t know respondents were grouped with No respondents. Unadjusted, crude ORs were calculated for associations between the exposure variables and outcome of interest. Effect modification between all possible combinations of exposure variables and stratification by community were evaluated, but no significant interactions were found and terms were excluded from further analysis. The behaviors of semicompliant persons (participated in MDA 1 3 times) were similar and did not vary statistically from fully compliant individuals. In the final analysis, semicompliant persons were excluded to best contrast the true differences between those who were compliant and those who were noncompliant. Polytomous logistic regression was not used in this study because fully compliant and semicompliant persons did not differ statistically with regard to exposure variables. A fully adjusted logistic regression model was developed that included all exposure variables. From this model, a final and parsimonious model was developed by removing terms that were statistically insignificant by Wald P value, or had large confidence intervals (CIs) that included the null. A final reduced model was selected and a chunk test was performed using log likelihood ratios to confirm that the terms removed were insignificant. Finally, multicollinearity between variables was assessed by examining condition indices and variance decomposition proportions (VDPs). Variables were considered for removal from analysis if they demonstrated condition indices > 30 or VDPs > 0.5. Data regarding antigen status was available for a limited number of persons in the cohort. A small sub-analysis was performed to compare the odds of infection in noncompliant persons with the odds of infection in compliant persons. RESULTS Study population. Of the 440 people selected for administration of the survey questionnaire, 367 (83.4%) were contacted and completed the study. A total of 73 (16.6%) persons could not be found, refused to participate, or were excluded from the study because they provided inadequate information for analysis. Of the respondents, 136 (37.1%) were classified as systematically noncompliant (cases) and 93 (25.3%) reported having participated in MDA all four years and were classified as controls. A total of 138 persons were semicompliant with MDA and excluded from the analysis. Preliminary analysis. Of all questions asked, unadjusted chi-square analysis showed that 22 of the exposure variables were associated with noncompliance (data not shown). The

3 COMPLIANCE WITH MDA FOR TREATMENT OF LF IN HAITI 285 age, sex, and education of compliant and noncompliant persons differed significantly by chi-square analysis. Noncompliance was strongly associated with questions concerning a person s perceived role in the community, as well as feelings that taking the pill did not personally benefit them in some way. A larger proportion of compliant people (72 of 93, 77.4%) answered Yes to feeling that it was important for persons to follow their example and take (or not take) pills; few noncompliant people indicated it was important (20 of 136, 14.7%). Of noncompliant persons, 85.3% (116/136) answered No/Don t know in response to this question (Table 1). Although chi-square analysis showed that questions related to the person s perceived risk of contracting LF were significantly associated with compliance (P 0.05), both cases and controls tended to answer that they did not know their risk level. This was especially true of the question could you have LF now? As shown in Table 1, 50.0% (46 of 92) of systemically compliant persons and 69.7% (92 of 132) of systematically noncompliant persons did not know if they were at risk for having LF. Similarly, compliant persons indicated they could swallow the pills during MDA (88 of 93, 94.6%), but noncompliant persons reported higher rates of either being unable to swallow the full complement of pills or not knowing if they could swallow the pills (61 of 136, 44.9%). Seventeen exposure variables seemed to be strongly associated with noncompliance and had ORs with CIs that did not contain the null (Table 1). Odds of noncompliance were high in persons who did not perceive their role in the community as important and could not or did not know if they could swallow the pills. Persons who answered no to the question would it be a good idea if people followed my example? were (95% CI ) times as likely to be noncompliant in unadjusted analysis. Similarly, odds of noncompliance were higher among those who could not swallow pills, or indicated they did not know if they could swallow the pills (OR 14.31, 95% confidence interval ). Odds of noncompliance were also high if people did not know the pills contained albendazole (OR 16.72; 95% CI ), which indicated possible lack of knowledge about one of the important collateral benefits of the program. Variables concerning perceived risk of contracting LF had wide CIs that included the null and did not appear to be significantly different amongst fully compliant and noncompliant persons. These variables were excluded from further analysis. Multivariate analysis. A fully adjusted logistic regression model was developed that included 22 exposure variables. The results of this model showed that many of the exposure variables had insignificant Wald P values and, in general, precision about point estimates was low. These results could be the result of the model containing too many predictor variables for our sample size. A parsimonious, reduced model was developed containing only the variables sex, would it be good if people followed my example and took or did not take the pill, do you think the LF program uses good drugs, have you ever heard of or seen the side effects of the pill, do you know that pills contain albendazole, have you heard about the LF program, could you swallow the pill, and are you protected by taking the pills even if others don t take them. Age and level of education were not found to be significant predictors of noncompliance in this group. In this population, the strongest predictors of compliance seem to be related to people s perceived role in community, the ability to take the pills, as well as overall knowledge of LF and the elimination program. The reduced model seemed to fit the data well both in terms of fit and precision, and the chunk test supported that the terms removed from the fully adjusted model were insignificant (P 0.545). Collinearity among variables was not found. The reduced model includes a relatively large number of predictor variables and maintains good precision around its parameter estimates (Table 2). DISCUSSION Noncompliant persons provide a potential reservoir for the filarial parasite and, in principle, could maintain transmission after cessation of MDA programs. As the prevalence of LF continues to decrease after interventions are implemented, the underlying factors driving noncompliance must be addressed to ensure the parasite is not harbored in clusters of people. In Haiti, knowledge-attitude-practice surveys administered during the 2000 MDA showed that age, sex, overall knowledge of LF, and awareness of the elimination program were significantly associated with noncompliance in MDA. 12 We conducted a case-control study to understand in greater detail why some persons were noncompliant after four years of MDA. This study identified several important predictors of noncompliance. We found that sex was the only demographic variable significantly associated with noncompliance, with women being 2.74 times more likely than men to be noncompliant (Table 2; 95% CI ). This result is similar to previous findings in Haiti showing that men were 2.97 times more likely to have taken filarial drugs (95% CI ). 12 Data from Pondicherry, India as well as other regions, demonstrate rates of noncompliance that are higher in men than women. 5,13 This discrepancy may be related to the Ministry of Health s initial programmatic decision to not treat women of reproductive age with albendazole to avoid inadvertent drug exposure during the first trimester of pregnancy. Unfortunately, this decision discouraged women in Leogane from participating in MDA. Although this decision was reversed by the third round of MDA, it is possible that women who did not participate in the MDA in 2000 failed to participate in later years. As the Leogane project continues, it is evident that public health officials need to take special care to emphasize the need to participate in MDA to women. Our findings continue to show that noncompliance is associated with poor knowledge of LF and the program to eliminate it. As shown in Table 2, people who did not participate in MDA were not likely to have known that the pills included albendazole (OR 5.08, 95% CI ) or to have heard about the program to eliminate LF (OR 3.17, 95% CI ). Furthermore, noncompliant persons also seem to indicate that it is not important if people follow their example (OR 5.74, 95% CI ). Persons who are not aware of the program to eliminate LF may lack a component of socialization that is present in compliant persons. Noncompliant persons may have either ignored the social mobilization campaign mounted by the program or simply never felt the need to participate. Because noncompliant persons also do not know the important collateral effects of

4 286 TALBOT AND OTHERS TABLE 1 Associations of questionnaire variables with noncompliance after four years of mass drug administration in Leogane, Haiti* Questionnaire variable Noncompliant Compliant Odds ratio (95% CI) Descriptive variables Age, years (17.0) 41 (17.9) 1.00 (Ref) (25.8) 35 (15.3) 1.77 ( ) (16.6) 17 (7.4) 2.35 ( ) Sex Male 47 (20.5) 46 (20.1) 1.00 (Ref) Female 89 (38.9) 47 (20.5) 1.85 ( ) Level of education Secondary or greater 43 (18.9) 53 (23.3) 1.00 (Ref) Primary 61 (26.9) 30 (13.2) 2.51 ( ) None 30 (13.2) 10 (4.4) 3.70 ( ) Role in community Does the choice to take the pills positively effect the community? Yes 24 (10.6) 43 (18.9) 1.00 (Ref) No 48 (21.1) 32 (14.1) 2.69 ( ) Don t know 62 (27.3) 18 (7.9) 6.17 ( ) Can LF be eliminated? Yes 62 (27.3) 69 (30.1) Ref No/don t know 74 (32.3) 24 (10.5) 3.43 ( ) Would it be good if people followed my example and took (or did not take) the pills? Yes 20 (8.7) 72 (31.4) 1.00 (Ref) No/don t know 116 (50.7) 21 (9.2) ( ) Trust in the healthcare system Does the hospital provide good care for SAE? Yes 102 (44.5) 83 (36.4) 1.00 (Ref) No/don t know 34 (14.8) 10 (4.4) 2.77 ( ) Do you think the LF program uses good drugs? Yes 75 (32.5) 86 (37.7) 1.00 (Ref) No/don t know 61 (26.9) 7 (3.1) 9.99 ( ) Do you think that LF staff discloses information well? Yes 68 (29.7) 67 (29.3) 1.00 (Ref) No/don t know 68 (29.7) 26 (11.4) 2.58 ( ) Fear of side effects/adverse events Ever heard/seen side effects of the pill? Yes 66 (28.9) 71 (31.1) 1.00 (Ref) No 69 (30.3) 22 (9.6) 3.37 ( ) Knowledge of disease and program to eliminate LF Knows that pills contain albendazole Yes 22 (9.6) 71 (31.0) 1.00 (Ref) No 114 (49.8) 22 (9.6) ( ) Heard about LF program? Yes 85 (37.3) 81 (35.5) 1.00 (Ref) No 50 (21.9) 12 (5.3) 3.97 ( ) Chance to question staff? Yes 9 (3.9) 18 (7.9) 1.00 (Ref) No 126 (55.3) 75 (32.9) 3.36 ( ) Access to MDA Can you swallow the pill? Yes 75 (32.8) 88 (38.6) 1.00 (Ref) No/don t know 61 (26.9) 5 (2.2) ( ) Are you concerned about the locality of the drug post? Yes 6 (2.6) 15 (6.6) 1.00 (Ref) No 130 (56.8) 78 (34.1) 4.17 ( ) Personal benefit in taking the pills Are you protected by taking the pills even if others don t take the pills? Yes 21 (9.4) 51 (22.8) 1.00 (Ref) No 37 (16.5) 30 (13.4) 3.00 ( ) Don t know 73 (32.6) 12 (5.4) ( ) Do pills have an immediate positive effect on your life? Yes 54 (23.6) 67 (29.3) 1.00 (Ref) No/don t know 82 (35.8) 26 (11.4) 3.91 ( ) Are pills the only protection from LF? Yes 38 (17.2) 51 (23.1) 1.00 (Ref) No 23 (10.4) 19 (8.6) 1.62 ( ) Don t know 68 (30.8) 22 (10.0) 4.15 ( ) Questionnaire variable TABLE 1 Continued Noncompliant Compliant Odds ratio (95% CI) If children take pills is there an immediate positive effect on their lives? Yes 56 (24.5) 73 (31.9) 1.00 (Ref) No/don t know 80 (34.9) 20 (8.7) 5.21 ( ) Risk of contracting LF Could you have LF now? Yes 8 (3.6) 14 (6.3) 1.00 (Ref) No 32 (14.3) 32 (14.3) 1.75 ( ) Don t know 92 (41.1) 46 (20.5) 3.50 ( ) Do you think you could get hydrocele or elephantiasis? Yes 33 (14.5) 28 (12.3) 1.00 (Ref) No 25 (11.0) 36 (15.8) 0.59 ( ) Don t know 77 (33.8) 29 (12.7) 2.25 ( ) Does everyone have an equal chance of contracting LF? Yes 58 (25.3) 48 (21.0) 1.00 (Ref) No/don t know 78 (34.1) 45 (19.7) 1.43 ( ) *CI confidence interval; Ref referent; LF lymphatic filariasis; SAE serious adverse event; MDA mass drug administration. Reference group. Statistically insignificant; null value (1) is contained in the CI. albendazole, 14 educational material that emphasizes this important role of MDA could provide an incentive for participation. Most persons who were noncompliant with MDA also indicated they did not know if they were protected by taking the pills, even if others did not take them (55.7%, 73 of 131; Table 1). This lack of knowledge was significantly associated with increased odds of noncompliance (OR 3.23, 95% CI ; Table 2). Additionally, many people who were noncompliant had never heard of or seen the adverse events (OR 2.11, 95% CI ) or indicated they felt the program did not use good drugs (OR 3.77, 95% CI ). Again, this may be a sign that compliant persons were more likely to have received instruction from health officials about the specific nature of LF and treatment, whereas noncompliant persons never received such information. Despite an apparent lack of fear or knowledge of adverse events, general wariness about MDA or the drugs themselves may still be an underlying cause of noncompliance. This mistrust may be a manifestation of the social upheaval that has characterized Haiti over the past two decades. Nonetheless, communication about the risks of infection and the benefits of treatment to the public should be improved. In adjusted regression models, several variables were not shown to be strong predictors of noncompliance, even though a significant association existed in crude analysis (Table 1). The apparent insignificance of the variables in fully adjusted logistic models could be explained by the proportion of persons who answered don t know to several questions. For example, 49.5% (46 of 93) of compliant persons didn t know if they could currently have LF. Similarly, 69.7% (92 of 132) of noncompliant persons didn t know. The percentages of people answering don t know tend to be high for questions that were not dichotomized. This finding could be indicative of a general lack of education concerning the nature of LF. However, these results could also suggest that the answer don t know is a surrogate for a don t care response. Future research should be directed toward the development of strategies to motivate noncompliant persons. Focus groups composed of noncompliant persons could provide a forum for understanding these issues.

5 COMPLIANCE WITH MDA FOR TREATMENT OF LF IN HAITI 287 TABLE 2 Reduced logistic regression model of associations of questionnaire variables with noncompliance after four years of mass drug administration in Leogane, Haiti* Questionnaire variable Noncompliant/compliant Odds ratio (95% CI) Wald P Descriptive variables Sex Male 47/ (Ref) Female 89/ ( ) Role in community Would it be good if people followed my example and took (or did not take) the pills? Yes 20/ (Ref) No/don t know 116/ ( ) < Trust in the healthcare system Do you think the LF program uses good drugs? Yes 75/ (Ref) No/don t know 61/ ( ) Fear of side effects/adverse events Ever heard/seen side effects of the pill? Yes 66/ (Ref) No 69/ ( ) Knowledge of disease and program to eliminate LF Knows that pills contain albendazole Yes 22/ (Ref) No 114/ ( ) Heard about LF program Yes 85/ (Ref) No 50/ ( ) Access to MDA Can you swallow the pill? Yes 75/ (Ref) No/don t know 61/ ( ) Personal benefit in taking the pills and risk of disease Are you protected by taking the pills even if others don t take the pills? Yes 21/ (Ref) No 37/ ( ) Don t know 73/ ( ) *CI confidence interval; Ref referent; LF lymphatic filariasis; MDA mass drug administration. Reference group 2 log likelihood statistic P for removed terms One programmatic obstacle for programs to eliminate LF is the inability for many persons to swallow the number of pills necessary to provide an adequate dosage. Adults in Leogane must swallow eight DEC tablets and take one chewable albendazole tablet. People who reported being unable to take the pill were 8.56 times as likely to be noncompliant as those who could (95% CI ). The difficulty in taking pills to prevent LF appears to be significant barrier for the MDA program in Leogane. If similar results are found in other areas, the global program should consider use of a chewable formulation of DEC or another delivery vehicle such as DEC-fortified salt. Furthermore, this finding has implications for integration of MDAs targeted at a broader spectrum of neglected tropical diseases, which could involve co-administration of drugs not only for LF and intestinal helminths, but also trachoma and schistosomiasis As part of a programmatic assessment, blood was collected for circulating filarial antigen testing (by immunochromatographic test [ICT]) to assess infection prevalence in all four communities. In Bire, Bonyotte, Guinebeau, and the Rue la Liberte neighborhood, ICT-positive rates were 30.0%, 3.5%, 34.5%, and 32.7%, respectively, and microfilaremia prevalence was 2.3%, 0.0%, 3.3%, and 2.9%, respectively (unpublished data). Antigen test results were available for 156 of 367 persons who completed the survey in this study (39 noncompliant, 62 semicompliant, and 55 compliant persons). Of these persons, 46 (29%) were antigen positive. Analysis of these data indicated that the odds of being antigen-positive were significantly higher in those persons who had never participated in MDA than in those who had ever participated in MDA (OR 2.35, 95% CI ). These data, although not definitive, suggest that noncompliance is a potential concern for LF elimination, at least in the Haitian context. The attitudes of persons who were semicompliant were generally similar to those who were fully compliant, and this may be a reflection of recall bias inherent in our study. It is also possible that persons who are semicompliant missed MDA because of circumstances unrelated to their knowledge of or attitudes about the program. Based on this interpretation, the MDA program has successfully targeted a large portion of the population. Because the odds of being ICT positive are significantly reduced for people who have ever participated in MDA, public health programs should strive to find ways to motivate noncompliant persons to participate in MDA at least once. This may entail healthcare workers personalizing drug distribution in those areas that are shown to have the highest levels of infection and the highest levels of noncompliance. Although data are not available from our study, future research should focus on determining if noncompliant persons are geographically clustered in communities, and if these persons are serving as foci of infection and ongoing transmission. If this is the case, programs to eliminate LF could intervene at different levels. For instance, integrated vector management in areas with clusters of persons who do not participate in MDA may be one way of preventing them from transmitting infection. Our study has several limitations. Responses to questions about participation in previous MDAs are subject to recall bias and because individual treatment records are not collected in Haiti, we have no way to verify compliance status. Although interviewers were not employees of either Hopital Ste. Croix or the LF program, it is possible that respondents provided answers based on their perception of what interviewers wanted to hear. In addition, by allowing proxies to provide information about compliance during the initial census, the numbers of cases and controls were reduced by misclassification. Finally, global positioning system data were collected as part of the census, but the information collected about compliance was not reliable enough to conduct a spatial analysis of the clustering of noncompliance. As programs to eliminate LF become more developed and have completed several rounds of MDA, it may become necessary to address the issue of systematic noncompliance through innovative public health approaches. It is imperative that as prevalence levels continue to decrease programs to eliminate LF become more effective in ensuring that proportions of persons who are systematically noncompliant are reduced. Designing interventions that reduce the number of noncompliant persons might make the difference between successful elimination of transmission of LF, and mere control of the parasite. Received July 6, Accepted for publication October 25, Acknowledgments: We thank the project staff in Leogane and the people living in the sentinel sites and Michael Deming for assistance with the sampling strategy for the case-control study.

6 288 TALBOT AND OTHERS Financial support: The study was supported by the Emerging Infections Program of CDC and by a grant from the Bill & Melinda Gates Foundation to the University of Notre Dame. Abigail Viall was supported by an Emerging Infectious Disease fellowship from the Association of Public Health Laboratories. Disclaimer: The views of the authors are their own and do not necessarily represent those of CDC. Authors addresses: Jeffrey T. Talbot, Center for Biostatistics and AIDS Research, Harvard University, 677 Huntington Avenue, Boston, MA Abigail Viall, Division of HIV/AIDS Prevention, Centers for Disease Control and Prevention, Mailstop D-21, 1600 Clifton Road, Atlanta, GA Abdel N. Direny and Madsen Beau de Rochars, Hopital Ste. Croix, Leogane, Haiti. David G. Addiss, Fetzer Institute, 9292 West KL Avenue, Kalamazoo, MI Thomas G. Streit, Center for Global Health, University of Notre Dame, 351 Galvin Hall, Notre Dame, IN Els Mathieu, Division of Parasitic Diseases Centers for Disease Control and Prevention, Mailstop F-22, 4770 Buford Highway, Atlanta, GA Patrick J. Lammie, Division of Parasitic Diseases Mailstop F-36, Centers for Disease Control and Prevention, 4770 Buford Highway, Atlanta, GA REFERENCES 1. Molyneux D, Lymphatic filariasis (elephantiasis) elimination: a public health success and development opportunity. Filaria J 2: World Health Organization, Global Programme to Eliminate Lymphatic Filariasis: Annual Report on Lymphatic Filariasis. Geneva: World Health Organization. 3. World Health Organization, Lymphatic filariasis. Wkly Epidemiol Rec 76: Michael D, Bundy DA, Grenfell BT, Re-assessing the global prevalence and distribution of lymphatic filariasis. Parasitology 112: Babu BV, Kar SK, Coverage, compliance and some operational issues of mass drug administration during the programme to eliminate lymphatic filariasis in Orissa, India. Trop Med Int Health 9: Ottesen EA, Duke BOL, Karam M, Behbehani K, Strategies and tools for the control/elimination of lymphatic filariasis. Bull World Health Organ 75: Stolk WA, Swaminathan S, van Oortmarssen GJ, Das PK, Habbema JD, Prospects for elimination of bancroftian filariasis by mass drug treatment in Pondicherry, India: a simulation study. J Infect Dis 188: de Rochars MF, Kanjilal S, Direny AN, Radday J, Lafontant JG, Mathieu E, Rheingans RD, Haddix AC, Streit TG, Beach MJ, Addiss DG, Lammie PJ, The Leogane, Haiti demonstration project: decreased microfilaremia and program costs after three years of mass drug administration. Am J Trop Med Hyg 73: Mathieu E, Direny AN, de Rochars MB, Streit TG, Addiss DG, Lammie PJ, Participation in three consecutive mass drug administrations in Leogane, Haiti. Am J Trop Med Hyg 11: Pichon G, Limitation and facilitation in the vectors and other aspects of the dynamics of filarial transmission: the need for vector control against Anopheles-transmitted filariasis. Ann Trop Med Parasitol 96: S143 S Esterre P, Plichart C, Sechan Y, Nguen L, The impact of 34 years of massive DEC chemotherapy on Wuchereria bancrofti infection and transmission: the Maupiti cohort. Trop Med Int Health 6: Mathieu E, Lammie PJ, Radday J, Beach MJ, Streit T, Wendt J, Addiss DG, Factors associated with participation in a campaign of mass treatment against lymphatic filariasis in Leogane, Haiti. Ann Trop Med Parasitol 98: Ramaiah KD, Das PK, Appavoo NC, Ramu K, Augustin DJ, Vijay Kurnar KN, Chandrakala AV, A program to eliminate lymphatic filariasis in Tamil Nadu state, India compliance with annual single-dose DEC mass treatment and some related operational aspects. Trop Med Int Health 5: de Rochars M, Direny AN, Roberts JM, Addiss DG, Radday J, Beach MJ, Streit TG, Dardith D, Lafontant JG, Lammie PJ, Community-wide reduction in prevalence and intensity of intestinal helminths as a collateral benefit of lymphatic filariasis elimination programs. Am J Trop Med Hyg 71: Brady MA, Hooper PJ, Ottesen EA, Projected benefits from integrating NTD programs in sub-saharan Africa. Trends Parasitol 22: World Health Organization, Preventive Chemotherapy in Human Helminthiasis: Coordinated Use of Anthelminthic Drugs in Control Interventions; A Manual for Health Professionals and Programme Managers. Geneva: World Health Organization.